Segment protection scheme for a network

Abstract

This invention broadly comprises a novel segment protection scheme (survivability framework) for a network, which we refer to as PROMISE (Protection using MultIple SEgments). It combines the best of existing link and path protection schemes (e.g., bandwidth efficiency and fast recovery). The PROMISE approach divides an active path or AP (along which a survivable connection is established) into several, possibly overlapping active segments or ASs, and then protects each AS with a detour called backup segment or BS (instead of protecting the AP as a whole as in path protection schemes). This facilitates the bandwidth sharing not only among the BSs for different APs, but also among those for the same AP. In addition, recovery time can be shortened due to the limited length of each AS and BS. This technology can be applied to MPLS, ATM, SONET, WDM and other high-speed link layers under the evolving G-MPLS framework.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application claims the benefit under 35 U.S.C. §119(e) of U.S. Provisional Application No. 60 / 380,491, filed May 14, 2002.FIELD OF THE INVENTION[0002]The present invention relates generally to computer and telecommunication networks and, more particularly, to a segment protection scheme for such networks which offers fast recovery, bandwidth efficiency, with path and link protection.BACKGROUND OF THE INVENTION[0003]As our national security, economy and even day-to-day life rely more and more on computer and telecommunication networks, avoiding prolonged disruptions to information exchange due to unexpected failures, such as a broken fiber link, becomes increasingly important. Hence, it is critical for a network to be survivable (or fault-tolerant).[0004]Two known survivability schemes are protection and restoration. The major difference between the two is that in protection, recovery from a failure (e.g., the detour set-up and spare ...

AI Technical Summary

Benefits of technology

[0027]This invention broadly comprises a novel segment protection scheme (survivability framework) for a network, which we refer to as PROMISE (Protection using MultIple SEgments). It combines the best of existing link and path protection schemes (e.g., bandwidth efficiency and fast recovery). A basic idea behind PROMISE is to divide an active path or AP (along which a survivable connection is established) into several, possibly overlapping active segments or ASs, and then protect each AS with a detour called backup segment or BS (instead of protecting the AP as a whole as in path protection schemes). This facilitates the bandwidth sharing not only among the BSs for different APs, but also among those for the same AP. In addition, recovery time can be shortened due to the limited length of each AS and BS. This technology can be applied to MPLS, ATM, SONET, WDM and other high-speed link layers under the evolving G-MPLS framework.

Problems solved by technology

In general, protection schemes can recover more quickly as the detour is already determined (as long as the detour is not broken), but are less bandwidth efficient than restoration schemes.

On the other hand, restoration schemes can survive one or multiple failures (as long as the destination is still reachable, and there is sufficient bandwidth), but they cannot guarantee the recovery time (including the time to find a detour), and / or the amount of information loss for real-time applications, making them unsuitable for mission-critical applications.

These issues are challenging because the objectives to optimize bandwidth usage, algorithm complexity and recovery time often conflict with each other.

However, finding an optimal pair of link (or node) disjoint paths that minimizes the total bandwidth consumption by the request, given that bandwidth sharing is possible, is an NP-hard problem.

In addition, the optimal pair, (if found, by using branch-and-bound techniques), often includes a long BP (consisting of many “zero or super low cost” links on which the bandwidth allocated to existing BPs can be shared), and this results in a long recovery time.

Existing efforts in achieving maximal bandwidth efficiency often resort to integer linear programming (ILP) which is not tractable / feasible for large-scale networks.

Other existing heuristic approaches sacrifice bandwidth efficiency, or trade recovery time for bandwidth efficiency.

The former chooses a set of rings that can cover all the nodes in a mesh network, but the traffic carried on any uncovered links in the mesh network cannot be protected against failure.

Even for a planar graph, it is difficult, if possible at all, to have small protection cycles (so that recovery can take place along shorter detours).

However, detours can also be long, and in addition, the number of p-cycles needed can be large, which requires complicated co-ordination amongst these p-cycles for the purpose of recovery.

More specifically, because it requires that the protection paths for all APs that terminate at a given egress router be determined simultaneously, one cannot take advantage of the bandwidth available somewhere else to support efficient dynamic establishment of connections.

This improves the bandwidth efficiency significantly, but results in longer detours and associated degradation of signal transmission quality.

In addition, the bandwidth efficiency of both approaches, especially the first, can be low due to the need to find a detour for each link, and neither approach, especially the second, made any attempts to limit the length of the detour.

Again, only rudimentary exhaustive search algorithms (with backtracking) and heuristics were suggested, and no performance results in terms of bandwidth efficiency, and recovery time (or the length of the detours) were provided.

But none of them contains, (or is supposed to contain), any implementation details such as algorithms or performance results, and it is clear that much work needs to be done in exploring the advantage of these protection schemes.

The bandwidth inefficiency problem is further exacerbated by the need to upgrade the bandwidth on all the links in a SDH / SONET ring (called the “fork-lift” requirement).

In addition, detours (loopbacked routes) can be very long, which not only wastes bandwidth, but also affects signal transmission performances such as signal-to-noise ratio (SNR) and bit-error-rate (BER), making all-optical data communications difficult.

On the other hand, link protection is less bandwidth efficient than path protection, B. Doshi and et al., “optical network design and restoration,”Bell Labs Technical Journal, pp.

As a variation of link or path protection, ring-based approaches generally are not bandwidth efficient.

In addition, they do not adapt well to the changes in the network topology due to their need to perform major reconstruction of the desired set of rings.

367-375, that they result in a significantly poor performance in terms of the ability to recover from subsequent failures.

As for existing non-ring based protection schemes, no existing schemes can achieve better bandwidth efficiency than shared path protection while having a much shorter backup segment, as well as a scalable algorithm.

Images